#http://stats.stackexchange.com/q/166592/1036

library(MASS)  #gung's answer using discriminant analysis
Iris  = iris[,c(1,5)]
model = lda(Species~Sepal.Length, Iris)
range(Iris$Sepal.Length)  # [1] 4.3 7.9
predData <- data.frame(Sepal.Length=seq(4, 8, .1))
predData$ldaClass <- predict(model,newdata=predData)$class

#effect plot, see http://www.jstatsoft.org/v32/i01/paper
library(effects)
library(nnet)

mul_mod <- multinom(Species~Sepal.Length, data=Iris)
eff <- effect("Sepal.Length", mul_mod)
plot(eff, style="stacked", rug = FALSE) #default plot(eff) shows the probabilities in small multiples
                                        #can make the same type of plot with posterior for LDA, or prob for rpart

predData$multinomClass <- predict(mul_mod, newdata=predData, type="class")

predData #exactly the same predictions with multinomial and lda

#with splines in prediction, stacked chart is basically the same
#library(splines)
#mul_modS <- multinom(Species~bs(Sepal.Length,3), data=Iris)
#effS <- effect("bs(Sepal.Length,3)", mul_modS)
#plot(effS, style="stacked", rug = FALSE)

#Placidia's option of using trees
library(rpart)
modelTree <- rpart(Species~Sepal.Length, Iris)
predData$TreeClass <- predict(modelTree,predData,type="class")

predData #only one difference with LDA or multinomial logistic regression